XAI-based Interference Classification in ISAC Systems Using RF Fingerprints

Interference detection and classification in integrated sensing and communication (ISAC) systems is a critical challenge for 6G networks, as it directly impacts system performance and reliability. In this paper, we aim to address this challenge by employing machine learning (ML) techniques, Random Forest, and XGBoost under varying signal-to-noise ratio (SNR) conditions by utilizing radio frequency fingerprints of the interference signal. Our results demonstrate that XGBoost outperforms Random Forest in terms of accuracy, macro-average, and weighted-average metrics as well as other classification metrics. To enhance the interpretability of ML models used, we leverage Explainable AI (XAI) tools, including SHapley Additive exPlanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME), to provide insight into the decision-making process of the models that are of paramount importance for AI applications in sixth-generation (6G) of wireless networks. These tools reveal the influence of individual features on the classification of each type of interference, offering a deeper understanding of the underlying patterns. By combining robust ML methods with XAI, this paper not only makes enhancements for interference classification in ISAC systems but also provides actionable insight for designing transceivers for ISAC systems, and tailored solutions for interference cancellation in ISAC systems, paving the way for more reliable and interpretable solutions in future 6G networks.